Formatting ggplots
Chenxin Li
2022-03-29
Codes are highlighted in grey.
Introduction
The ggplot package in R is a powerful plotting tool.
It’s easy to use, but you can make complex plots with rather simple syntax.
“gg” stands for graphical grammar -
It means different layers of objects and elements are layered on top of the previous to generate the plot.
You can find the ggplot cheatsheet
This workbook will go over different features of ggplot
- geoms
- scales
- guides
- themes
- facets
…and how to make the clearest, prettiest ggplot possible.
Load packages
You might need to install some of these packages
library(ggplot2)
library(tidyr)
library(dplyr)##
## Attaching package: 'dplyr'## The following objects are masked from 'package:stats':
##
## filter, lag## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionWe’ll use the R built-in dataset ToothGrowth.
This dataset is about different doses of vitamin C on tooth growth of guinea pigs
Ref: Crampton, E. W . (1947) . The Journal of Nutrition, 33(5), 491–504 . doi: 10.1093/jn/33.5.491 .
head(ToothGrowth)str(ToothGrowth)## 'data.frame': 60 obs. of 3 variables:
## $ len : num 4.2 11.5 7.3 5.8 6.4 10 11.2 11.2 5.2 7 ...
## $ supp: Factor w/ 2 levels "OJ","VC": 2 2 2 2 2 2 2 2 2 2 ...
## $ dose: num 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 ...It has 60 rows and 3 columns.
Two types of supp: OJ and VC
Looks like len is tooth length and the dependent variable.
Let’s make the most basic plot fir st
Coordinate system
Dose on X, length on Y, color as supp
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
geom_point(aes(color = supp)) 
I prefer to type out the data frame first, then pipe into ggplot(). It’s more readable this way.
Thus ToothGrowth %>% ggplot(...).
The line ggplot(aes(x = ..., y = ...)) specifies what are the x and y axis.
In this case does on x axis, len on y axis.
Then you add a new layer (in this case points) by using +
and in a new line you write geom_poits(). “Geom” stands for geommetric object.
In this case geom_points() produces points.
Inside geom_point() you want to color by supp, so you write aes(color = supp).
This is not bad… Now let’s use a custom palette called “wisteria”. It’s just a set of color names.
Adjust geom color/fill
wisteria <- c("grey65", "burlywood3", "khaki2", "plum1", "lightcyan2", "cornflowerblue", "slateblue3")This is just a list of color names in R.
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
geom_point(aes(color = supp)) +
scale_color_manual(values = wisteria[c(1, 7)]) #calling the 1st and 7th color 
You can customize color using scale_color_manual(),
in side which, you specifies the names of the colors,
in this case the color names are pre-specified by a vector called wisteria.
Change axis and legend labels - labs()
Now I want to change the axis and legend titles. How do I do that?
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
geom_point(aes(color = supp)) +
scale_color_manual(values = wisteria[c(1, 7)]) +
labs(x = "dose", #the labs option allow you to change axis and legend labels.
y = "teeth length",
color = "supplement")
The labs() function allows you to change axis and legend labels.
Should be self-explanatory.
You can see the dots are overlapping each other. How do I make them spread out a bit horizontally?
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
geom_point(aes(color = supp),
position = position_jitter(0.1), #horizontal spread of 0.1
alpha = 0.8, #20% transparent. When dots do overlap, we can tell
size = 3) + #make them larger and easier to see
scale_color_manual(values = wisteria[c(1, 7)]) +
labs(x = "dose",
y = "teeth length",
color = "supplement")
There are a few options in geom_points. The position = position_jitter(0.1) argument adds random horizontal spread of 0.1 to each dot.
The alpha = 0.8 makes your dots 20% transparent. (alpha = 1 is 0% transparent and default.)
The size option allows you to make dots larger or smaller .
This is great, but I can make it even better by adding border to each dots. How to do that?
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
geom_point(aes(fill = supp), #note that change to fill = supp
position = position_jitter(0.1),
alpha = 0.8,
size = 3,
shape = 21, #shape = 21 gives you open circle
color = "black") + #black border around each dot
scale_fill_manual(values = wisteria[c(1, 7)]) + #change to scale_fill here
labs(x = "dose",
y = "teeth length",
fill = "supplement") #change to fill = here
Color vs fill
Well there are more options in geom_point. One of them is shape. geom_point allows different shapes, such as square, cross and etc.
My favorite is open circle. To use the open circles, you call shape = 21.
The advantage of open circle is that it has both color and fill properties.
To help you understand the difference between color and fill, let’s compare the following two plot s.
data.frame(
x = c("group1", "group2"),
y = c(10, 20)
) %>%
ggplot(aes(x = x, y = y)) +
geom_bar(stat = "identity", aes(fill = x)) +
ggtitle("fill by x")
data.frame(
x = c("group1", "group2"),
y = c(10, 20)
) %>%
ggplot(aes(x = x, y = y)) +
geom_bar(stat = "identity", aes(color = x)) +
ggtitle("color by x")
I hope you can see the difference. If we use fill, the entire interior is colored, or filled. If we use color, only the outline is colored.
Coming back to using open circles in geom_point(),
Open circles takes both fill and color.
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
geom_point(aes(fill = supp), #note that change to fill = supp
position = position_jitter(0.1),
alpha = 0.8,
size = 3,
shape = 21, #shape = 21 gives you open circle
color = "black") + #black border around each dot
scale_fill_manual(values = wisteria[c(1, 7)]) + #change to scale_fill here
labs(x = "dose",
y = "teeth length",
fill = "supplement") #change to fill = here
In this case, I use color = "black" to make black outlines for each dot.
Then I used aes(fill = supp) to fill the interior of dots by the supp.
Now, to specify colors, I have to use scale_fill_manual(),
instead of scale_color_manual(), because the variable supp is now reflected by fill, not colo r.
But there is still a problem…
Note that when you use jitter, the horizontal spread is completely random.
This means every time you run the code, the position of the dots are different.
How do I fix tha t?
You can add a seed argument within position_jitter(), that will make it reproduci ble
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666), #added seed = 666 here
#doesn't matter which seed number you use
#as long as it's the same number the pattern will be the same
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
labs(x = "dose",
y = "teeth length",
fill = "supplement")
In this case, I used seed = 666.
But you can use any seed number. As long as it’s the same number, the plot will look the same every time you run the code.
Stat_summary - very powerful visualization tool
Let’s say I want a curve for each supplement, what should I do?
Your first reaction is geom_line, no?
Unfortunately, that’s not the right answer.
geom_line will try to connect each dot,
but what we want is try to connect the means of each dose for each supp.
We’ll use stat_summary() ins tead
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
stat_summary(geom = "line", # draw a line
fun = mean, # summarise by averaging, thus mean
aes(group = supp, color = supp), # one line per supp, color by supp
size = 1.2) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
labs(x = "dose",
y = "teeth length",
fill = "supplement")
In stat_summary(), I put geom = "line", which asked it to draw lines.
Then I put fun.y = mean, which asked it summarize y values using the mean() function.
Then I put aes(group = supp, color = supp), which asked for one line per supp,
and color those lines by sup p.
Well, there’s a problem. The lines are not the same color as the dots.
That’s because we specified the interior colors of dots using scale_fill_manual(),
but we never specified the color of the lines.
It’s easy to fi x. We will use scale_color_manual() to specify colors for the lines to o.
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) + #add scale_color_manual here
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement") #change the color legend title here 
Better!
Note that since the title and scales of fill and color are the same, ggplot merged them in the legend.
Now I want to add errorbars to the line, how do I do that?
We’ll use stat_summary again,
but this time instead of geom = "line", we’ll do geom = "errorbar"
Besides geom = "errorbar", another option is geom = "linerange"
linerange is errorbar without the horizontal b ars
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "errorbar", #draw errorbars (another option is geom = "linerange")
fun.data = mean_se, #mean_se specifies the errorbar to be based on standard error (se)
aes(group = supp), #one errorbar per dose per supp
width = 0.1) + #make the bar narrower
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement") 
Now there are error bars.
Here is another option to show the errorbar using stat_summary(geom = "ribbon")
I like ribbon better . You can see for yourself .
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "ribbon", #draw ribbons around line
fun.data = mean_se,
aes(group = supp,
fill = supp),
alpha = 0.5) + #make ribbons more transparent
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement")
I like the ribbon better.
It looks very nice for high transparency with alpha = 0.5.
But feel free to use errorbar or linerange if you prefer thos e
Adjust legends and axis
Note that in the experiment, there’s never a dose = 1.5.
We can turn off the x axis tick mark of 1.5 by using scale_x
(You can adjust y axis using scale_y as well, if you ever need to )
BTW, if you ever need to log10 transform axis, you can do scale_x_log10() or scale_y_log10()
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "ribbon",
fun.data = mean_se,
aes(group = supp,
fill = supp),
alpha = 0.5) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_continuous(breaks = c(0.5, 1, 2)) + #use scale_x_continuos here because x is continuos
#if you have discrete x, you will need scale_x_discrete
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement")
Because x axis value is numeric and continuous, you will use scale_x_continuous()
And inside, you set the break makers, in this case 0.5, 1 and 2.
Now the x = 1.5 mark is turned off .
I realized the color and fill scales are redundant, so we can turn off one of them.
To do so, we’ll use guides.
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "ribbon",
fun.data = mean_se,
aes(group = supp,
fill = supp),
alpha = 0.5) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_continuous(breaks = c(0.5, 1, 2)) +
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") #turn off color legend using color = "none" 
The guides() function can do a lot of stuff. Here I turned color legend off by putting color = "none" in guides().
Adjust themes
The default theme is nice, but it’s not the best
You can use the theme() function to build your own theme
Let’s remove all the theme element first using theme_minimal()
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "ribbon",
fun.data = mean_se,
aes(group = supp,
fill = supp),
alpha = 0.5) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_continuous(breaks = c(0.5, 1, 2)) +
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
theme_minimal() 
Theme minimal is the minimalist theme. It turns off most theme elements.
But you can add elements that you want back.
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "ribbon",
fun.data = mean_se,
aes(group = supp,
fill = supp),
alpha = 0.5) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_continuous(breaks = c(0.5, 1, 2)) +
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
theme_minimal() +
theme(axis.line = element_line(size = 1.2)) + #add axis line back
theme(text = element_text(size = 12, color = "black", face = "bold")) + #make text clearer and darker
theme(axis.text = element_text(size = 12, color = "black", face = "bold")) 
The theme() function takes a lot of options, such as axis.line, text and all that.
You can look up the options here
The rule of thumb is if you want to adjust lines, such as axis line, then you use element_line()
Then you specify color and size in element_line()
If you want to adjust text, then you use element_text(),
again in which you specify color and size in element_text().
The white background and black lines really put the data forward and emphasize the data
Now notice the legend is on the right and it is taking up space?
You can move it into the plot using theme(legend.position = )
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "ribbon",
fun.data = mean_se,
aes(group = supp,
fill = supp),
alpha = 0.5) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_continuous(breaks = c(0.5, 1, 2)) +
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
theme_minimal() +
theme(legend.position = "bottom") + #move legend to bottom
theme(axis.line = element_line(size = 1.2)) +
theme(text = element_text(size = 12, color = "black", face = "bold")) +
theme(axis.text = element_text(size = 12, color = "black", face = "bold")) 
In theme(), you can put legend.position = "bottom".
You can also put "top", "right" or "left".
The more fancy thing is to move legend inside the graph area. How to do that?
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "ribbon",
fun.data = mean_se,
aes(group = supp,
fill = supp),
alpha = 0.5) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_continuous(breaks = c(0.5, 1, 2)) +
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
theme_minimal() +
theme(legend.position = c(0.8, 0.2)) + #this is a tricky command, and you have to manually adjust it
#c(1, 1) means upper right corner of the graph area,
#c(1, 0) means the lower right corner of the graph area
theme(axis.line = element_line(size = 1.2)) +
theme(text = element_text(size = 12, color = "black", face = "bold")) +
theme(axis.text = element_text(size = 12, color = "black", face = "bold"))
In legend.position = c(x, y), you manually adjust where the legend will show up.
For example c(1, 1) means upper right corner of the graph area,
and c(1, 0) means the lower right corner of the graph are a
Now the legend is in the graph area.
This is nice because now the legend doesn’t take up extra space
One last adjustment you can do is change the number of rows and column of the legend.
Now we have two rows, one column. What if I want one row and two columns?
We’ll use guides again.
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "ribbon",
fun.data = mean_se,
aes(group = supp,
fill = supp),
alpha = 0.5) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_continuous(breaks = c(0.5, 1, 2)) +
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
guides(fill = guide_legend(nrow = 1, ncol = 2)) + # set fill legend to 1 row 2 columns.
theme_minimal() +
theme(legend.position = c(0.8, 0.2)) +
theme(axis.line = element_line(size = 1.2)) +
theme(text = element_text(size = 12, color = "black", face = "bold")) +
theme(axis.text = element_text(size = 12, color = "black", face = "bold"))
This line guides(fill = guide_legend(nrow = N)) is very useful we you have limited space and a lot of legend items
It also works as guides(color = guide_legend(nrow = N)) if you need to arrange color legends. Again, think about the difference between fill and color.
Faceting
faceting is very important when you have a factorial design.
In this experiment, we have supp as a factor, and dose as another factor.
So this is a factorial design
faceting allows you and readers to focus on the true comparison of intere st
Let’s say we are interested in comparing the effect of each supp across different doses…
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
facet_grid(. ~ supp) + #see comments after plot for explanations
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "ribbon",
fun.data = mean_se,
aes(group = supp,
fill = supp),
alpha = 0.5) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_continuous(breaks = c(0.5, 1, 2)) +
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
guides(fill = guide_legend(nrow = 1, ncol = 2)) +
theme_minimal() +
theme(legend.position = c(0.8, 0.2)) +
theme(axis.line = element_line(size = 1.2)) +
theme(text = element_text(size = 12, color = "black", face = "bold")) +
theme(axis.text = element_text(size = 12, color = "black", face = "bold"))
Now it makes two plots.
The facet syntax works like this:
facet_grid(row ~ column) We are using supp as columns, and only 1 row, so facet_grid(. ~ supp). The . before the ~ is just a space holder .
You can increase the panel spacing too
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
facet_grid(. ~ supp) +
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "ribbon",
fun.data = mean_se,
aes(group = supp,
fill = supp),
alpha = 0.5) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_continuous(breaks = c(0.5, 1, 2)) +
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
guides(fill = guide_legend(nrow = 1, ncol = 2)) +
theme_minimal() +
theme(legend.position = "none") + #this time the legend is redundant w/ the panel label, so I turn it off.
theme(axis.line = element_line(size = 1.2)) +
theme(text = element_text(size = 12, color = "black", face = "bold")) +
theme(axis.text = element_text(size = 12, color = "black", face = "bold")) +
theme(panel.spacing = unit(1.2, "lines")) #make 1.2 lines of spacing between the subplots (panels)
Panel spacing was increased using theme(panel.spacing = unit(1.2, "lines"))
Note that this time the legend is redundant with the panel label,
so I turn it off using legend.position = "none" in theme()
Great !
But now say I am actually interested in comparing between supp at each dose, what do I do?
We’ll need facet_grid(. ~ dose), no?
ToothGrowth %>%
ggplot(aes(x = dose, y = len)) +
facet_grid(. ~ dose) + #change to ~ dose
stat_summary(geom = "line",
fun = mean,
aes(group = supp, color = supp),
size = 1.2) +
stat_summary(geom = "ribbon",
fun.data = mean_se,
aes(group = supp,
fill = supp),
alpha = 0.5) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_continuous(breaks = c(0.5, 1, 2)) +
labs(x = "dose",
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
guides(fill = guide_legend(nrow = 1, ncol = 2)) +
theme_minimal() +
theme(legend.position = "none") +
theme(axis.line = element_line(size = 1.2)) +
theme(text = element_text(size = 12, color = "black", face = "bold")) +
theme(axis.text = element_text(size = 12, color = "black", face = "bold")) +
theme(panel.spacing = unit(1.2, "lines")) ## geom_path: Each group consists of only one observation. Do you need to adjust
## the group aesthetic?
## geom_path: Each group consists of only one observation. Do you need to adjust
## the group aesthetic?
## geom_path: Each group consists of only one observation. Do you need to adjust
## the group aesthetic?
Now there is a problem. The lines and ribbons are gone.
That’s because in each panel, there is only one dose. There is nothing to connect.
And the x axis is redundant with the panel.
So what we really need is to put supp on x axis inste ad
ToothGrowth %>%
ggplot(aes(x = supp, y = len)) +
facet_grid(. ~ dose) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
#scale_x_continuous(breaks = c(0.5, 1, 2)) + remove this. this time x axis is discrete
labs(x = "supplement", #now x = supplment
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
guides(fill = guide_legend(nrow = 1, ncol = 2)) +
theme_minimal() +
theme(legend.position = "none") + #this time the legend is redundant with the x label, so I turn it off.
theme(axis.line = element_line(size = 1.2)) +
theme(text = element_text(size = 12, color = "black", face = "bold")) +
theme(axis.text = element_text(size = 12, color = "black", face = "bold")) +
theme(panel.spacing = unit(1.2, "lines")) 
There is no numeric trend over the x variable, so I removed the ribbon and line.
But I can still add the errorbars back.
ToothGrowth %>%
ggplot(aes(x = supp, y = len)) +
facet_grid(. ~ dose) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
stat_summary(geom = "errorbar", #draw errorbars
width = 0.2, size = 1,
aes(group = supp), #one bar per supp per dose
fun.data = mean_se) + #errorbar base on standard error (SE)
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
labs(x = "supplement", #now x = supplment
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
guides(fill = guide_legend(nrow = 1, ncol = 2)) +
theme_minimal() +
theme(legend.position = "none") +
theme(axis.line = element_line(size = 1.2)) +
theme(text = element_text(size = 12, color = "black", face = "bold")) +
theme(axis.text = element_text(size = 12, color = "black", face = "bold")) +
theme(panel.spacing = unit(1.2, "lines")) 
Pretty good.
There is one problem. “OJ” and “VC” text appear multiple times.
We can turn them off and only show the legend .
ToothGrowth %>%
ggplot(aes(x = supp, y = len)) +
facet_grid(. ~ dose) +
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
stat_summary(geom = "errorbar",
width = 0.2, size = 1,
aes(group = supp),
fun.data = mean_se) +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_discrete(labels = NULL) + #turn off x axis text using labels = NULL
labs(x = "supplement",
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
guides(fill = guide_legend(nrow = 1, ncol = 2)) +
theme_minimal() +
theme(legend.position = "bottom") + #put the legends back
theme(axis.line = element_line(size = 1.2)) +
theme(text = element_text(size = 12, color = "black", face = "bold")) +
theme(axis.text = element_text(size = 12, color = "black", face = "bold")) +
theme(panel.spacing = unit(1.2, "lines"))
Now just a couple final touches. Note that the text “supplement” appears twice, and dose appears none.
We can put the panel labels to the bottom, and switch the x axis label to dose instead.
ToothGrowth %>%
ggplot(aes(x = supp, y = len)) +
facet_grid(. ~ dose, switch = "x") + #switch = "x" puts x panel labels to the bottom instead of top
geom_point(aes(fill = supp),
position = position_jitter(0.1, seed = 666),
alpha = 0.8,
size = 3,
shape = 21,
color = "black") +
stat_summary(geom = "errorbar",
width = 0.2, size = 1,
aes(group = supp),
fun.data = mean_se) +
scale_fill_manual(values = wisteria[c(1, 7)]) +
scale_color_manual(values = wisteria[c(1, 7)]) +
scale_x_discrete(labels = NULL) +
labs(x = "dose", #now x is dose again
y = "teeth length",
fill = "supplement",
color = "supplement") +
guides(color = "none") +
guides(fill = guide_legend(nrow = 1, ncol = 2)) +
theme_minimal() +
theme(legend.position = "top") + #legends to the top, make space for x axis label
theme(axis.line = element_line(size = 1.2)) +
theme(text = element_text(size = 12, color = "black", face = "bold")) +
theme(axis.text = element_text(size = 12, color = "black", face = "bold")) +
theme(panel.spacing = unit(1.2, "lines")) +
theme(strip.placement = "outside") # put the text 0.5, 1, and 2 outside the axis line
Now it really looks like dose is x axis, but in fact it is the faceting factor.
Congratulation! You have produced a publication quality plot!
I think this is the best plot we can make for faceting by dose.
The best practice of plotting is always idenify the comparison of interest,
remove redundant and unnecessary elements,
try to make it as pretty as you can, and
be clear on what you want to sh ow.
Exercise - Now time for you to try
Let’s use the R built-in dataset CO2 as example
Ref: Potvin, C., Lechowicz, M. J. and Tardif, S. (1990), Ecology, 71, 1389–1400.
head(CO2)str(CO2)## Classes 'nfnGroupedData', 'nfGroupedData', 'groupedData' and 'data.frame': 84 obs. of 5 variables:
## $ Plant : Ord.factor w/ 12 levels "Qn1"<"Qn2"<"Qn3"<..: 1 1 1 1 1 1 1 2 2 2 ...
## $ Type : Factor w/ 2 levels "Quebec","Mississippi": 1 1 1 1 1 1 1 1 1 1 ...
## $ Treatment: Factor w/ 2 levels "nonchilled","chilled": 1 1 1 1 1 1 1 1 1 1 ...
## $ conc : num 95 175 250 350 500 675 1000 95 175 250 ...
## $ uptake : num 16 30.4 34.8 37.2 35.3 39.2 39.7 13.6 27.3 37.1 ...
## - attr(*, "formula")=Class 'formula' language uptake ~ conc | Plant
## .. ..- attr(*, ".Environment")=<environment: R_EmptyEnv>
## - attr(*, "outer")=Class 'formula' language ~Treatment * Type
## .. ..- attr(*, ".Environment")=<environment: R_EmptyEnv>
## - attr(*, "labels")=List of 2
## ..$ x: chr "Ambient carbon dioxide concentration"
## ..$ y: chr "CO2 uptake rate"
## - attr(*, "units")=List of 2
## ..$ x: chr "(uL/L)"
## ..$ y: chr "(umol/m^2 s)"Task: produce a plot that includes the info for Type and Treatment, with uptake on Y axis.
Let’s say we are interested in the effect of “Treatment” in each Type.
Hint:
- What should be on the x axis?
- What variable do you color or fill with?
- What faceting layout do you want to use?
Use the format:
CO2 %>%
ggplot(aes(x = ???, y = uptake)) +
facet_grid(??? ~ ???) +
geom_ponit(aes()) +
stat_summary(geom = errorbar) +
scale_color_manual() +
scale_fill_manual() +
scal e_…_…. +
labs() +
guides() +
theme_minimal()+
theme(legend.position = …) +
theme(axis.line = element_line(size = 1.2)) +
theme(text = element_text(size = 12, color = “black”, face = “bold”)) +
theme(axis.text = element_text(size = 12, color = “black”, fa ce = “bold”))
Now try to make the best plot you can!